APPLIED PHYSICS LETTERS VOLUME 85, NUMBER 10 6 SEPTEMBER 2004

High-frequency surface acoustic wave device based on thin-film

piezoelectric interdigital transducers
A. K. Sarin Kumar,a) P. Paruch, and J.-M. Trisconeb)
DPMC, University of Geneva, 24 Quai E. Ansermet, 1211 Geneva 4, Switzerland
W. Daniau and S. Ballandras
Laboratoire de Physique et Métrologie des Oscillateurs, CNRS UPR3203, 32 Avenue de l’Observatoire,
F25044, Besançon, France
L. Pellegrino and D. Marré
INFM, Research unit of Genoa, Dodecaneso 33, 16146 Genova, Italy
T. Tybell
Department of Electronics and Telecommunications, Norwegian University of Science and Technology, 7491
Trondheim, Norway
(Received 21 April 2004; accepted 9 July 2004)
Using high-quality epitaxial c-axis Pb共Zr0.2Ti0.8兲O3 films grown by off-axis magnetron sputtering
onto metallic (001) Nb-doped SrTiO3 substrates, a nonconventional thin-film surface acoustic wave
device based on periodic piezoelectric transducers was realized. The piezoelectric transducers
consist of a series of ferroelectric domains with alternating polarization states. The artificial
modification of the ferroelectric domain structure is performed by using an atomic force microscope
tip as a source of electric field, allowing local switching of the polarization. Devices with 1.2 and
0.8 ␮m wavelength, defined by the modulation period of the polarization, and corresponding to
central frequencies in the range 1.50– 3.50 GHz have been realized and tested. © 2004 American
Institute of Physics. [DOI: 10.1063/1.1787897]

Surface acoustic waves (SAW) are used for filtering ap-
plications in mobile communication and for passive signal
processing devices, such as delay lines and resonators.1 It is
expected that the frequency range allocated for personal me-
dia communications will increase into the 5 – 10 GHz range,2
requiring high-frequency SAW filters.3 High-frequency SAW
devices would not only be useful in future applications, but
could also be of interest in the study of liquid-solid inter-
faces, or electron-phonon interactions.4–7
In classical SAW devices, periodic metallic interdigital
transducers (IDTs), deposited on uniformly polarized piezo-
electric crystals or films, act as electric input and output
ports.8 The application of an appropriate radio frequency (rf)
signal to the IDT produces a deformation of the material
surface and results in the launching of SAW, which travel
along the piezoelectric surface with a phase velocity given
by the physical properties of the material. These surface
waves can then be converted back to an electric signal by a
receiving IDT. The center frequency 共f c兲 of the SAW device
is given by f c = vp / ␭, where the wavelength ␭ is the distance
between the IDT fingers of the same electrode, and v p is the
phase velocity in the material. In order to increase f c, either
the wavelength of the device has to be decreased or a mate-
rial with a higher phase velocity has to be used as the trav-
eling medium. Typical phase velocities of most piezoelectric
materials are between 2500– 6000 m / s resulting in working
frequencies of 1 − 3 GHz when standard photolithography
(minimum finger width 0.5 ␮m) is used for defining the
IDTs. Prototype SAW filters in the GHz range have also been
demonstrated by growing the piezoelectric materials on sub-
a)
Present address: Department of Electonics and Telecommunications, Nor-
wegian University of Science and Technology,7491 Trondheim, Norway.
b)
Author to whom correspondence should be addressed; electronic mail:
Jean-Marc.Triscone@Physics.UniGE.ch
strates with high phase velocities like diamond or
sapphire.9–11 Another approach to reaching higher working
frequencies is to fabricate devices with smaller wavelengths
by decreasing the distance between the IDT fingers, for in-
stance by using electron beam lithography. With this tech-
nique, prototype devices with wavelengths smaller than
1 ␮m have been fabricated, leading to center frequencies of
up to 20 GHz.11,12
In this letter we report on the modeling, fabrication, and
electrical measurements of prototype SAW devices based on
a nonconventional transducer principle. We show that nano-
scale ferroelectric domain manipulation and atomic force mi-
croscopy (AFM) writing of periodic piezoelectric transducers
in epitaxial films is a promising approach, allowing the
implementation of very high-frequency SAW devices.
In this original device, the conventional metallic IDT has
been replaced by a piezoelectric interdigital transducer (PIT)
consisting of oppositely polarized ferroelectric domains,
schematically illustrated in Fig. 1(a). Such a transducer is
fabricated by writing line-shaped domains with alternating
polarization on high-quality c-axis Pb共Zr0.2Ti0.8兲O3 (PZT)
films grown on metallic (001) Nb:SrTiO3 substrates.13 The
local polarization control is achieved by using the metallic
tip of an AFM as a local electric-field source, applying a
voltage between the tip and the conducting substrate (used as
the ground electrode).14,15 In previous studies on similar
films, it was shown that domains with radii as small as 20 nm
can be written down to 10 nm apart using AFM.15 This con-
trol directly determines the precision at which the PIT can be
written and the minimum wavelength of the device. To excite
the PIT, a single electrode is deposited on its surface [see
Fig. 1(a)], in contrast to standard SAW devices, and a rf
voltage is then applied between the top electrode and the
conducting substrate. Since the sign of the piezoelectric co-
efficient depends on the polarization direction, the applied rf

0003-6951/2004/85(10)/1757/3/$22.00 1757 © 2004 American Institute of Physics

Downloaded 04 Jun 2009 to 129.241.108.137. Redistribution subject to AIP license or copyright; see http://apl.aip.org/apl/copyright.jsp
1758 Appl. Phys. Lett., Vol. 85, No. 10, 6 September 2004
FIG. 1. A schematic diagram of the SAW device with PITs (piezoelectric
IDTs). The PITs are made of ferroelectric domains with alternating polar-
ization states written in the c-axis oriented PZT films. (b) Schematic repre-
sentation of the domain deformation with applied rf voltage. The deforma-
tion (elongation or contraction) for a given voltage depends on the
polarization direction.
voltage will produce a deformation (elongation or contrac-
tion) of the domains, as shown in Fig. 1(b), launching a
surface wave when ␭f c = v p. This mechanical deformation of
the material is complex, with one end of the domains being
“clamped” to the substrate, which is assumed to be rigid and
the other end being essentially free to deform (although cov-
ered by the electrode). The wavelength ␭ of the device is the
distance between the domains of the same polarization, con-
trary to standard IDT-based SAW devices for which the
wavelength equals twice the mechanical period of the grat-
ings. Consequently, for any given resolution, the operating
frequency of PIT-based SAW devices is expected to be twice
that of a standard SAW device. Furthermore, the size of the
domains can be controlled by AFM and thus made very
small. These are the main innovations of the proposed ap-
proach, which can lead to the fabrication of very high-
frequency SAW filters, possibly above 20 GHz.
To further understand the behavior of such a structure,
numerical simulations have been performed.16 In these simu-
lations, the elementary cell of the periodic array is composed
of a 0.2-␮m-thick PbZr0.5Ti0.5O3 layer deposited on a (001)
SrTiO3 substrate (considered electrically conductive) with a
Pt electrode above the piezoelectric layer. The polarization is
assumed to exhibit a 1.2 ␮m period, which is the wavelength
of one of the devices described below. The harmonic admit-
tance is then computed assuming an infinite periodic struc-
ture excited in phase. Four coupled modes are found at
1.57 GHz 共v p = 1884 m / s , Ks2 = 14.5%兲, 2.02 GHz 共v p
= 2424 m / s , Ks2 = 2.4%兲, 2.96 GHz 共v p = 3552 m / s , Ks2
= 1.8%兲 and 3.90 GHz 共v p = 4680 m / s , Ks = 7.2%兲, where Ks2
2
is the electromechanical coupling coefficient. An interesting
result of the simulation is that the wave radiation losses into
bulk started appearing only around 4 GHz, in contrast to
conventional devices where bulk radiation occurs around
2 GHz. This suggests that bulk acoustic energy radiation loss
is reduced in this transducer geometry as compared to con-
ventional SAW devices.
To experimentally test the PIT principle, c-axis-oriented
epitaxial PZT films ⬃0.2 ␮m thick were fabricated on (001)
Nb:SrTiO3 (0.2 at. wt. % Nb) by off-axis rf magnetron sput-
tering. The details of the preparation and characterization can
be found in Ref. 17. X-ray diffraction analysis reveals only
Downloaded 04 Jun 2009 to 129.241.108.137. Redistribution subject to AIP license or copyright; see http://apl.aip.org/apl/copyright.jsp
Sarin Kumar et al.
FIG. 2. Piezoelectric microscopy image of one of the prototype SAW de-
vices.The distance between the input and output PIT is around 70 ␮m and
that between two adjacent line-shaped domains of opposite polarization is
0.6 ␮m, leading to a 1.2 ␮m wavelength. (b) Line section through piezore-
sponse (PR) image of one of the PITs.
c-axis reflections with FWHM (full width at half maximum)
values of 0.05 and 0.15° for the (001) and (002) peaks, re-
spectively. Topographic measurements of the sample sur-
faces using AFM revealed smooth surfaces with a root-mean-
square roughness of typically 3 Å. To fabricate the PITs, a
150⫻ 100 ␮m2 area was first defined photolithographically
on the film.18 An area of around 130⫻ 65 ␮m2 was then
uniformly polarized by slowly scanning the AFM tip while
applying a constant −12 V between the tip and the conduct-
ing substrate. This prepoling is performed to avoid any
domain-wall scattering of the surface waves traveling be-
tween the PITs. The input and output PITs were then fabri-
cated by writing line-shaped domains with alternating polar-
ization inside a 20⫻ 20 ␮m2 area by alternately applying a +
and −12 V bias to the tip. The accuracy of the domain width
is typically ±0.08– 0.1 ␮m. Subsequently, Pt electrodes with
a thickness of 20 nm were sputtered on the PITs. Figure 2(a)
shows a piezoelectric microscopy image of a prototype SAW
device. Here, the distance between the input and output PITs
is around 70 ␮m and that between two adjacent line-shaped
domains of opposite polarization is 0.6 ␮m, leading to a
wavelength of 1.2 ␮m. Figure 2(b) shows a line section
through the piezoresponse image of one of the PITs. It is
interesting to note that the size of this SAW device is typi-
cally 100⫻ 100 ␮m2. Based on ladder architecture, one can
expect further size reduction using PIT-based structures in
comparison with standard IDT-based devices for similar
wavelengths.9
One of the critical issues regarding this device is the
stability of the written ferroelectric domains. Heat treatment
(during photolithographic processes) and the applied electric
field during device operation could lead to instabilities of the
artificially written domain structure. For the former case it
has been found that the domains are stable for more than
three months (the duration of the measurements), even after
heating to 250° C and heavy photolithographic processing.
To estimate the effect of the electric field on domain-wall
Appl. Phys. Lett., Vol. 85, No. 10, 6 September 2004
FIG. 3. Reflection 共S11兲 spectra of 1.2 ␮m (dotted line) and 0.8 ␮m (solid
line) wavelength devices with PITs.
motion, we calculated the maximum field applied to the de-
vice (about 2 ⫻ 104 V / cm, below the 105 V / cm PZT coer-
cive field) to get the domain-wall velocity.19 A velocity of
about ⬃10−11 m / s was found, leading to negligible domain-
wall motion for the high working frequencies used here.
Electrical measurements were carried out using an An-
ritsu ® network analyzer and a set of Picoprobe ® coplanar
probes.20 Since our substrate is metallic and also partly be-
cause of the high PZT losses, a significant damping of the
SAW is observed in this prototype device, not allowing di-
rect SAW transmission measurements to be performed. For
this reason, we focus here on the S11 reflection measurements
which allow us to demonstrate the PIT-operation. Figure 3
shows the reflection measurement 共S11兲 results of the PIT-
based SAW devices with different wavelength in the fre-
quency range of 3 GHz. For a 1.2 ␮m device (dotted line),
two reflection minima at 1.50 and 2.02 GHz are observed, in
good agreement with the first two modes (1.57 and
2.02 GHz) observed in the simulation results described
above. In a first-order approximation (i.e., neglecting the dis-
persion behavior), a change in wavelength will produce a
corresponding change in the working frequency of the de-
vice. For example, a wavelength change from 1.2 to 0.8 ␮m
would correspond to a frequency shift of 1.5. To test this, we
fabricated another device with a wavelength of 0.8 ␮m. The
result for the 0.8 ␮m device is shown in Fig. 3 (solid line).
Two reflection minima close to 2.57 and 3.44 GHz can be
seen, corresponding to a frequency shift of 1.7 with respect
to the 1.2 ␮m device. This small discrepancy in frequency
ratio from the expected value of 1.5 is not unreasonable in a
first-order approximation, and taking into account the writing
errors in the linewidth of each domain, ±0.08– 0.1 ␮m.
These results are an experimental demonstration related to
the launching of a SAW using PITs.
With the above mentioned experimental observations,
one can conclude that the PITs operate in accordance with
the basic SAW equation, f c = v p / ␭, and in agreement with the
simulations. According to theoretical analysis, the reflection
minima observed experimentally in the vicinity of 2.02 GHz
共␭ = 1.2 ␮m兲 and 3.44 GHz 共␭ = 0.8 ␮m兲 will correspond to
the vibration mode shown in Fig.1(b). However, the contri-
bution close to 1.5 GHz 共␭ = 1.2 ␮m兲 and 2.57 GHz 共␭
= 0.8 ␮m兲 is less intuitive and is the equivalent of an anti-
symmetric Lamb mode of the PZT layer. A small contribu-
tion is also observed at 3.4 GHz 共␭ = 1.2 ␮m兲, for which a
match with the simulation is more difficult to establish. The
reflection curves reported in Fig. 3 also show damped con-
tributions, whereas sharper peaks are expected from theoret-
ical analyses, which include only losses due to acoustic ra-
diation in the substrate. This suggests that other sources of
Downloaded 04 Jun 2009 to 129.241.108.137. Redistribution subject to AIP license or copyright; see http://apl.aip.org/apl/copyright.jsp
Sarin Kumar et al. 1759
losses have to be considered in the computations. For in-
stance, PZT has performed poorly in rf applications21–23 and
may thus be a source of important losses. Also, conduction
currents due to the wave penetration may be generated in the
substrate yielding additional losses.24 More experimental
studies are in progress to address these issues.
In conclusion, we have proposed an original transducer
concept for rf SAW devices based on a periodic distribution
of oppositely poled piezoelectric domains and succeeded in
fabricating prototypes of such devices on epitaxial PZT films
prepared on metallic Nb-SrTiO3 substrates. The measured
S11 parameter for the two devices with different wavelengths
exhibited a clear shift in frequency response in agreement
with the basic principle of SAW devices.
This work was supported by the Swiss National Science
Foundation through the National Centre of Competence in
Research “Materials with Novel Electronic Properties-
MaNEP” and Division II, NEDO, and ESF (Thiox).
1
C. C. W. Ruppel, R. Dill, A. Fischerauer, G. Fischerauer, W. Gawlik, J.
Machui, F. Muller, L. Reindl, W. Ruile, G. Scholl, I. Schropp, and K. Ch.
Wagner, IEEE Trans. Ultrason. Ferroelectr. Freq. Control 40, 438 (1993).
2
K. M. Larkin, G. R. Kline, and T. McCarron, IEEE Trans. Microwave
Theory Tech. 43, 2933 (1995).
3
S.-H. Lee, K. H. Yoon, and J.-K. Lee, Appl. Phys. Lett. 92, 4062 (2002),
and references therein.
4
R. W. Morse and H. V. Bohm, Phys. Rev. 108, 1094 (1957).
5
D. J. Bishop, C. M. Varma, B. Batlogg, and E. Bucher, Phys. Rev. Lett.
53, 1009 (1984).
6
M. Saint-Paul, F. Pourtier, B. Pannetier, J. C. Villegier, and R. Nava,
Physica C 183, 257 (1991).
7
S.–G. Lee, C.–C. Chi, G. Koren, and A. Gupta, Phys. Rev. B 43, 5459
(1991).
8
R. M. White and F. W. Voltmer, Appl. Phys. Lett. 7, 314 (1965).
9
C. K. Campbell, Surface Acoustic Wave Devices for Mobile and Wireless
Communications (Academic, San Diego, 1998).
10
H. Nakahata, K. Higaki, A. Hachigo, S. Shikata, N. Fujimori, Y. Taka-
hashi, T. Kajihara, and Y. Yamamoto, Jpn. J. Appl. Phys. 33, 324 (1994).
11
Y. Takagaki, P. V. Sanots, E. Wiebicke, O. Brandt, H.-P. Schönherr, and K.
12
H. Ploog, Appl. Phys. Lett. 81, 2538 (2002).
S. Lehtonen, J. Koskela, M. M. Salomaa, V. P. Plessky, M. Honkanen, and
J. Turunen, Appl. Phys. Lett. 75, 142 (1999).
13
A. K. Sarin Kumar, P. Paruch, D. Marré, L. Pellegrino, T. Tybell, S.
Ballandras, and J.-M. Triscone, Integr. Ferroelectr. 63, 55–62 (2004); D.
Marré, T. Tybell, C. Beneduce, and J.-M. Triscone (to be published).
14
C. H. Ahn, T. Tybell, L. Antognazza, K. Char, R. H. Hammond, M. R.
Beasley, Ø Fisher, and J.-M. Triscone, Science 276, 1100 (1997).
15
P. Paruch, T. Tybell, and J.-M. Triscone, Appl. Phys. Lett. 79, 530 (2001).
16
S. Ballandras, V. Laude, Th. Pastureaud, M. Wilm, W. Daniau, A. Rein-
hardt, W. Steichen, and R. Lardat, Proc.-IEEE Ultrason. Symp. 1, 321
(2002).
17
T. Tybell, C. H. Ahn, and J.-M. Triscone, Appl. Phys. Lett. 75, 856
(1999).
18
Amorphous, 250-nm-thick CeO2 were deposited everywhere on PZT films
except in the defined 150⫻ 100 ␮ m2, allowing us to reduce the capaci-
tance of the device.
19
T. Tybell, P. Paruch, T. Giamarchi, and J.-M. Triscone, Phys. Rev. Lett.
89, 097601 (2002).
20
Before the transmission measurements, we calibrated the network analyzer
21
using an impedance standard substrate.
Q.-X. Su, P. Kirby, E. Komuro, M. Imura, Q. Zhang, R. Whatmore, IEEE
Trans. Ultrason. Ferroelectr. Freq. Control 49-4, 769 (2001).
22
T. Omori, H. Makita, M. Takamatsu, K. Y. Hashimoto, and M. Yamaguchi,
Proc.-IEEE Ultrason. Symp. 2, 995 (1999).
23
M. Yamaguchi, K. Hashimoto, R. Nanjo, H. Hanazawa, S. Tsutsumi, and
T. Yonezawa, Proc. IEEE-IFCS Frequency Control Symposium, 1, 544
(1997).
24
M. Solal, Th. Pastureaud, S. Ballandras, B. Aspar, B. Biasse, W. Daniau,
J. M. Hode, S. Calisti, and V. Laude, Proc.-IEEE Ultrason. Symp. 1, 131
(2002).